After a few minutes of silence, Karim Akhlal turned around. The young man in his twenties had entered the employment agency looking for a part-time job, but the staff had ignored him since his arrival. Feeling uneasy, he closed the door behind him.
Although difficult to prove, Karim Akhlal could guess the reason for the way he was treated that day. As a Moroccan Dutchman, he had often had to deal with discrimination on the labor market. Sometimes because of his appearance (tinted, black beard), sometimes the mention of his name was enough for a rejection, in his own words.
Outside the employment agency, Akhlal had decided: I’m going to show how recruitment and selection of personnel should be done. Now, more than fifteen years later, the now 39-year-old Akhlal has his own office, located in a converted library in Hoofddorp. With his start-up Recrout, founded in 2015, he wants to make the labor market ‘fairer’. Issues such as ‘origin, color and smell’ should no longer play a role in the recruitment and selection of personnel. Akhlal wants to achieve that with the help of artificial intelligence (AI).
Behind the chair in Akhlal’s office are drawings of a fruit tree and a house, drawn by his 7-year-old daughter Nora. Its staff – nine employees in total – are spread all over the world. “India, Mauritius, Ukraine…”, Akhlal sums up. He has never seen some of the software developers in real life, only through the screen.
The software that Recrout sells revolves around determining the competencies of applicants. Job seekers are presented with a questionnaire, which results in scores that determine the extent to which someone has an aptitude for certain skills such as leadership and concentration. The computer then predicts which candidate best fits the vacancy that a company has.
Recrout is certainly not the only Dutch company that offers AI technology for recruitment and selection procedures. For example, there is TextKernel, a larger player in the recruitment tech market, with over 200 employees and offices in the Netherlands, Germany, France and the US. The company uses deep learning, where the computer learns to recognize patterns in resumes and can easily put the information in a large file. This gives a recruiter a quick overview of possible candidates. In addition, according to director Gerard Mulder, the software is trained “to understand what is written”, even if someone expresses themselves with other words of similar meaning, or even in a different language. For example, it no longer matters whether someone registers as a nurse or nurse nurse to be filtered by the computer.
Or take Recruitee and RecruitNow, two companies that launched their first AI tools in recent months that recruiters can use to write job descriptions. There are also parties such as the American HireVue, which makes video software that can be used to assess a candidate’s verbal and non-verbal reactions during the interview. The employer then receives the most suitable candidate on a silver platter from the computer.
‘booming’
The emergence of technology and self-learning algorithms in application procedures is therefore not new. However, this development is accelerating, Akhlal sees. “AI is booming. We need to be aware that it’s not going away and think about how we can use it to our advantage while staying in control.” According to Janneke Oostrom, professor of work and organizational psychology at Tilburg University, the use of AI “certainly” improves efficiency in the recruitment and selection of personnel. For example, systems are able to quickly process a large bulk of application data and tools such as HireVue allow candidates to record their video application at any time of the day and be immediately assessed.
But will applying for jobs really become fairer through the use of AI? That’s hard to say. “There is still little evidence for this,” says Oostrom. This is primarily due to the fact that little research has been done on it. “In the past, the focus was mainly on what the jobseeker experiences during a job application procedure that uses technology, and not whether it also works to predict future work behavior.”
According to Oostrom, applicants get a sense of injustice more quickly in such an application procedure than when a person goes through the resumes and conducts the interview. “You don’t know how you are judged. With a human application procedure you have that idea. While it is precisely known that people have prejudices – conscious and unconscious.”
The second point is that not all companies are able to explain exactly how their technology works. “I understand that they claim that their algorithm predicts something,” says Oostrom. “But how is it formed and built up? This is often so complex that even the recruiter who starts working with the technology, or the candidate who applies for a job, has no idea.” She believes that companies would do well to submit the selection software to the independent Netherlands Test Matters Committee (COTAN), which can provide a definite answer about the quality.
Read also: How artificial intelligence is putting 300 million jobs at risk
When artificial intelligence is given free rein without it being clear on what choices are based, things can go very wrong. For example, Amazon scrapped its AI application tool in 2015 because it had taught itself that male candidates are more suitable than women. Video recruitment company HireVue, which claims to be able to predict future work behavior based on non-verbal reactions, was The Washington Post called “pseudoscience” by experts. Not only would it be doubtful whether facial expressions could be used to say anything about someone’s skills, the technology would also disadvantage nervous people, people who spoke in a language other than their mother tongue or (visually) disabled people. The company dismissed the criticism as “uninformed” in a response.
In his office, Recrout director Akhlal calls concerns about the lack of transparency about the technology “justifiable”. The tests he uses have been approved by the COTAN mentioned by Oostrom. The algorithm has not (yet) been approved – “an external audit is indeed a to-do for us”.
If it’s up to Akhlal, AI won’t take over the entire job from the recruiter. “As an employer, you should not blindly rely on the results. You should see it as a first selection tool, after which a company can enter into a conversation with a candidate who has been found suitable by the system. The applicant has already passed the first selection moment at which a prejudice can arise about the color of the skin or the headscarf.”
‘Call Honk’
Gerard Mulder of Textkernel also says he is careful about using AI in his technology. He touches on another concern about self-learning algorithms: in the EU, companies are bound by strict data privacy legislation, which raises the question of whether a company has enough data from applicants to train the model sufficiently. “It is often recruiters who incorporate their own ideas into the algorithm, so that human biases can still occur.”
Due to the increasing attention for artificial intelligence, Akhlal also sees new parties “hooking” that they develop recruitment technology with the help of AI. “Investors love it,” he says. According to him, existing Dutch tech companies are well aware that there are dark sides to AI, but a “rat races. We have to stay sober, what is really relevant for the goal we are aiming for and what is a gimmick?” According to Akhlal, it now becomes “even more important” that companies such as Recrout can prove to the outside world what they say.
Mulder is not afraid that a proliferation of AI recruitment tech will have a negative effect on his company, because excesses may occur more often than at Amazon. “Of course, the limit of what AI can do is still unknown. It is also becoming easier to build tools using existing software such as ChatGPT. But before recruitment tech is actually used, it must first comply with many rules, including those of organizations themselves.”
“Artificial intelligence certainly has the potential to make the labor market fairer,” thinks Professor Oostrom. When and how AI succeeds, “continues to speculate”. According to her, a bill that has already been passed that requires employers to record the working method of recruitment and selection in writing to promote equal opportunities can bring about a “major change” in the selection world. “Recruitment tech companies will soon also have to be able to demonstrate that their software prevents discrimination.”
Akhlal is convinced that he can. Although he also signs for a discrimination-free labor market without Recrout having developed the decisive tech for this. “If my daughter Nora soon enters a fair labor market and I have contributed to it, my mission will have been successful.”
A version of this article also appeared in the newspaper of May 23, 2023.